I've used multiple GLP-1s. The weight came off. So did things I didn't plan to lose: muscle, hair, mineral status, sleep depth, libido. This is the tool I wish I'd had on day one of my first 2.5 mg. Drop in your dose, your training, your protein, and watch the rest of your body respond.
I built each of these models because I had the question and couldn't find an honest answer. Not from the prescriber. Not from the subreddit. Not from the AI summary that just regurgitated the prescriber.
Up to 38% of weight lost on a GLP-1 is lean mass when nothing is done about it. Drop your dose, your training, and your real protein intake into the lab and it tells you which dose week you're catabolic — before your DEXA does.
Nausea, sulfur burps, constipation, gallbladder pressure, hair shed. The model maps probability against dose, titration speed, and what you actually ate this week. Predict bad days before you live them.
75% of people regain most of the weight within a year of stopping. Test-fly the off-ramp — slow taper, protein-led nutrition, resistance training — as a 12-month scenario, not a "maintenance" platitude.
Long-term effects on bone density, gastric motility, thyroid C-cells. The lab doesn't pretend to have answers. It surfaces the confidence band, the studies it pulled from, and where you're extrapolating into uncertainty.
One real scenario someone built in the closed beta. Not a recommendation. A worked example — to show you the shape of what you'd be designing for yourself.
Not every marker on a Function Health panel matters for what you're doing. These are the ones that do, with the reason next to each.
The harm engine is hard-coded, not statistical. MAOIs + SSRIs, methylene blue + serotonergics, contraindicated peptide stacks, dosing into a known drug-interaction window — the lab refuses to render the prediction and surfaces peer-reviewed warnings instead.
It will also flag when you're outside the population the literature was sampled from — pregnant, breastfeeding, under 18, or stacking peptides or supplements the model has low confidence in. You always know what the model knows and what it's guessing.
Every protocol you save can be exported as a clean PDF: mechanisms cited, doses listed, comparator scenario attached, predicted markers tracked. The point isn't to replace your prescriber. The point is to walk into a 12-minute appointment with a structured artifact instead of a story.
This is not medical advice. By design. It is education that makes you a more informed person sitting across from someone who can give it.
Closed beta. Tell us you're building a routine for glp-1 users and we'll get you in. Every founder@ email gets read.